1
2	How to Get Your Change Into the Linux Kernel
3		or
4	Care And Operation Of Your Linus Torvalds
5
6
7
8For a person or company who wishes to submit a change to the Linux
9kernel, the process can sometimes be daunting if you're not familiar
10with "the system."  This text is a collection of suggestions which
11can greatly increase the chances of your change being accepted.
12
13This document contains a large number of suggestions in a relatively terse
14format.  For detailed information on how the kernel development process
15works, see Documentation/development-process.  Also, read
16Documentation/SubmitChecklist for a list of items to check before
17submitting code.  If you are submitting a driver, also read
18Documentation/SubmittingDrivers; for device tree binding patches, read
19Documentation/devicetree/bindings/submitting-patches.txt.
20
21Many of these steps describe the default behavior of the git version
22control system; if you use git to prepare your patches, you'll find much
23of the mechanical work done for you, though you'll still need to prepare
24and document a sensible set of patches.  In general, use of git will make
25your life as a kernel developer easier.
26
27--------------------------------------------
28SECTION 1 - CREATING AND SENDING YOUR CHANGE
29--------------------------------------------
30
31
320) Obtain a current source tree
33-------------------------------
34
35If you do not have a repository with the current kernel source handy, use
36git to obtain one.  You'll want to start with the mainline repository,
37which can be grabbed with:
38
39  git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
40
41Note, however, that you may not want to develop against the mainline tree
42directly.  Most subsystem maintainers run their own trees and want to see
43patches prepared against those trees.  See the "T:" entry for the subsystem
44in the MAINTAINERS file to find that tree, or simply ask the maintainer if
45the tree is not listed there.
46
47It is still possible to download kernel releases via tarballs (as described
48in the next section), but that is the hard way to do kernel development.
49
501) "diff -up"
51------------
52
53If you must generate your patches by hand, use "diff -up" or "diff -uprN"
54to create patches.  Git generates patches in this form by default; if
55you're using git, you can skip this section entirely.
56
57All changes to the Linux kernel occur in the form of patches, as
58generated by diff(1).  When creating your patch, make sure to create it
59in "unified diff" format, as supplied by the '-u' argument to diff(1).
60Also, please use the '-p' argument which shows which C function each
61change is in - that makes the resultant diff a lot easier to read.
62Patches should be based in the root kernel source directory,
63not in any lower subdirectory.
64
65To create a patch for a single file, it is often sufficient to do:
66
67	SRCTREE= linux
68	MYFILE=  drivers/net/mydriver.c
69
70	cd $SRCTREE
71	cp $MYFILE $MYFILE.orig
72	vi $MYFILE	# make your change
73	cd ..
74	diff -up $SRCTREE/$MYFILE{.orig,} > /tmp/patch
75
76To create a patch for multiple files, you should unpack a "vanilla",
77or unmodified kernel source tree, and generate a diff against your
78own source tree.  For example:
79
80	MYSRC= /devel/linux
81
82	tar xvfz linux-3.19.tar.gz
83	mv linux-3.19 linux-3.19-vanilla
84	diff -uprN -X linux-3.19-vanilla/Documentation/dontdiff \
85		linux-3.19-vanilla $MYSRC > /tmp/patch
86
87"dontdiff" is a list of files which are generated by the kernel during
88the build process, and should be ignored in any diff(1)-generated
89patch.
90
91Make sure your patch does not include any extra files which do not
92belong in a patch submission.  Make sure to review your patch -after-
93generating it with diff(1), to ensure accuracy.
94
95If your changes produce a lot of deltas, you need to split them into
96individual patches which modify things in logical stages; see section
97#3.  This will facilitate review by other kernel developers,
98very important if you want your patch accepted.
99
100If you're using git, "git rebase -i" can help you with this process.  If
101you're not using git, quilt <http://savannah.nongnu.org/projects/quilt>
102is another popular alternative.
103
104
105
1062) Describe your changes.
107-------------------------
108
109Describe your problem.  Whether your patch is a one-line bug fix or
1105000 lines of a new feature, there must be an underlying problem that
111motivated you to do this work.  Convince the reviewer that there is a
112problem worth fixing and that it makes sense for them to read past the
113first paragraph.
114
115Describe user-visible impact.  Straight up crashes and lockups are
116pretty convincing, but not all bugs are that blatant.  Even if the
117problem was spotted during code review, describe the impact you think
118it can have on users.  Keep in mind that the majority of Linux
119installations run kernels from secondary stable trees or
120vendor/product-specific trees that cherry-pick only specific patches
121from upstream, so include anything that could help route your change
122downstream: provoking circumstances, excerpts from dmesg, crash
123descriptions, performance regressions, latency spikes, lockups, etc.
124
125Quantify optimizations and trade-offs.  If you claim improvements in
126performance, memory consumption, stack footprint, or binary size,
127include numbers that back them up.  But also describe non-obvious
128costs.  Optimizations usually aren't free but trade-offs between CPU,
129memory, and readability; or, when it comes to heuristics, between
130different workloads.  Describe the expected downsides of your
131optimization so that the reviewer can weigh costs against benefits.
132
133Once the problem is established, describe what you are actually doing
134about it in technical detail.  It's important to describe the change
135in plain English for the reviewer to verify that the code is behaving
136as you intend it to.
137
138The maintainer will thank you if you write your patch description in a
139form which can be easily pulled into Linux's source code management
140system, git, as a "commit log".  See #15, below.
141
142Solve only one problem per patch.  If your description starts to get
143long, that's a sign that you probably need to split up your patch.
144See #3, next.
145
146When you submit or resubmit a patch or patch series, include the
147complete patch description and justification for it.  Don't just
148say that this is version N of the patch (series).  Don't expect the
149subsystem maintainer to refer back to earlier patch versions or referenced
150URLs to find the patch description and put that into the patch.
151I.e., the patch (series) and its description should be self-contained.
152This benefits both the maintainers and reviewers.  Some reviewers
153probably didn't even receive earlier versions of the patch.
154
155Describe your changes in imperative mood, e.g. "make xyzzy do frotz"
156instead of "[This patch] makes xyzzy do frotz" or "[I] changed xyzzy
157to do frotz", as if you are giving orders to the codebase to change
158its behaviour.
159
160If the patch fixes a logged bug entry, refer to that bug entry by
161number and URL.  If the patch follows from a mailing list discussion,
162give a URL to the mailing list archive; use the https://lkml.kernel.org/
163redirector with a Message-Id, to ensure that the links cannot become
164stale.
165
166However, try to make your explanation understandable without external
167resources.  In addition to giving a URL to a mailing list archive or
168bug, summarize the relevant points of the discussion that led to the
169patch as submitted.
170
171If you want to refer to a specific commit, don't just refer to the
172SHA-1 ID of the commit. Please also include the oneline summary of
173the commit, to make it easier for reviewers to know what it is about.
174Example:
175
176	Commit e21d2170f36602ae2708 ("video: remove unnecessary
177	platform_set_drvdata()") removed the unnecessary
178	platform_set_drvdata(), but left the variable "dev" unused,
179	delete it.
180
181You should also be sure to use at least the first twelve characters of the
182SHA-1 ID.  The kernel repository holds a *lot* of objects, making
183collisions with shorter IDs a real possibility.  Bear in mind that, even if
184there is no collision with your six-character ID now, that condition may
185change five years from now.
186
187If your patch fixes a bug in a specific commit, e.g. you found an issue using
188git-bisect, please use the 'Fixes:' tag with the first 12 characters of the
189SHA-1 ID, and the one line summary.  For example:
190
191	Fixes: e21d2170f366 ("video: remove unnecessary platform_set_drvdata()")
192
193The following git-config settings can be used to add a pretty format for
194outputting the above style in the git log or git show commands
195
196	[core]
197		abbrev = 12
198	[pretty]
199		fixes = Fixes: %h (\"%s\")
200
2013) Separate your changes.
202-------------------------
203
204Separate each _logical change_ into a separate patch.
205
206For example, if your changes include both bug fixes and performance
207enhancements for a single driver, separate those changes into two
208or more patches.  If your changes include an API update, and a new
209driver which uses that new API, separate those into two patches.
210
211On the other hand, if you make a single change to numerous files,
212group those changes into a single patch.  Thus a single logical change
213is contained within a single patch.
214
215The point to remember is that each patch should make an easily understood
216change that can be verified by reviewers.  Each patch should be justifiable
217on its own merits.
218
219If one patch depends on another patch in order for a change to be
220complete, that is OK.  Simply note "this patch depends on patch X"
221in your patch description.
222
223When dividing your change into a series of patches, take special care to
224ensure that the kernel builds and runs properly after each patch in the
225series.  Developers using "git bisect" to track down a problem can end up
226splitting your patch series at any point; they will not thank you if you
227introduce bugs in the middle.
228
229If you cannot condense your patch set into a smaller set of patches,
230then only post say 15 or so at a time and wait for review and integration.
231
232
233
2344) Style-check your changes.
235----------------------------
236
237Check your patch for basic style violations, details of which can be
238found in Documentation/CodingStyle.  Failure to do so simply wastes
239the reviewers time and will get your patch rejected, probably
240without even being read.
241
242One significant exception is when moving code from one file to
243another -- in this case you should not modify the moved code at all in
244the same patch which moves it.  This clearly delineates the act of
245moving the code and your changes.  This greatly aids review of the
246actual differences and allows tools to better track the history of
247the code itself.
248
249Check your patches with the patch style checker prior to submission
250(scripts/checkpatch.pl).  Note, though, that the style checker should be
251viewed as a guide, not as a replacement for human judgment.  If your code
252looks better with a violation then its probably best left alone.
253
254The checker reports at three levels:
255 - ERROR: things that are very likely to be wrong
256 - WARNING: things requiring careful review
257 - CHECK: things requiring thought
258
259You should be able to justify all violations that remain in your
260patch.
261
262
2635) Select the recipients for your patch.
264----------------------------------------
265
266You should always copy the appropriate subsystem maintainer(s) on any patch
267to code that they maintain; look through the MAINTAINERS file and the
268source code revision history to see who those maintainers are.  The
269script scripts/get_maintainer.pl can be very useful at this step.  If you
270cannot find a maintainer for the subsystem you are working on, Andrew
271Morton (akpm@linux-foundation.org) serves as a maintainer of last resort.
272
273You should also normally choose at least one mailing list to receive a copy
274of your patch set.  linux-kernel@vger.kernel.org functions as a list of
275last resort, but the volume on that list has caused a number of developers
276to tune it out.  Look in the MAINTAINERS file for a subsystem-specific
277list; your patch will probably get more attention there.  Please do not
278spam unrelated lists, though.
279
280Many kernel-related lists are hosted on vger.kernel.org; you can find a
281list of them at http://vger.kernel.org/vger-lists.html.  There are
282kernel-related lists hosted elsewhere as well, though.
283
284Do not send more than 15 patches at once to the vger mailing lists!!!
285
286Linus Torvalds is the final arbiter of all changes accepted into the
287Linux kernel.  His e-mail address is <torvalds@linux-foundation.org>.
288He gets a lot of e-mail, and, at this point, very few patches go through
289Linus directly, so typically you should do your best to -avoid-
290sending him e-mail.
291
292If you have a patch that fixes an exploitable security bug, send that patch
293to security@kernel.org.  For severe bugs, a short embargo may be considered
294to allow distributors to get the patch out to users; in such cases,
295obviously, the patch should not be sent to any public lists.
296
297Patches that fix a severe bug in a released kernel should be directed
298toward the stable maintainers by putting a line like this:
299
300  Cc: stable@vger.kernel.org
301
302into the sign-off area of your patch (note, NOT an email recipient).  You
303should also read Documentation/stable_kernel_rules.txt in addition to this
304file.
305
306Note, however, that some subsystem maintainers want to come to their own
307conclusions on which patches should go to the stable trees.  The networking
308maintainer, in particular, would rather not see individual developers
309adding lines like the above to their patches.
310
311If changes affect userland-kernel interfaces, please send the MAN-PAGES
312maintainer (as listed in the MAINTAINERS file) a man-pages patch, or at
313least a notification of the change, so that some information makes its way
314into the manual pages.  User-space API changes should also be copied to
315linux-api@vger.kernel.org.
316
317For small patches you may want to CC the Trivial Patch Monkey
318trivial@kernel.org which collects "trivial" patches. Have a look
319into the MAINTAINERS file for its current manager.
320Trivial patches must qualify for one of the following rules:
321 Spelling fixes in documentation
322 Spelling fixes for errors which could break grep(1)
323 Warning fixes (cluttering with useless warnings is bad)
324 Compilation fixes (only if they are actually correct)
325 Runtime fixes (only if they actually fix things)
326 Removing use of deprecated functions/macros
327 Contact detail and documentation fixes
328 Non-portable code replaced by portable code (even in arch-specific,
329 since people copy, as long as it's trivial)
330 Any fix by the author/maintainer of the file (ie. patch monkey
331 in re-transmission mode)
332
333
334
3356) No MIME, no links, no compression, no attachments.  Just plain text.
336-----------------------------------------------------------------------
337
338Linus and other kernel developers need to be able to read and comment
339on the changes you are submitting.  It is important for a kernel
340developer to be able to "quote" your changes, using standard e-mail
341tools, so that they may comment on specific portions of your code.
342
343For this reason, all patches should be submitted by e-mail "inline".
344WARNING:  Be wary of your editor's word-wrap corrupting your patch,
345if you choose to cut-n-paste your patch.
346
347Do not attach the patch as a MIME attachment, compressed or not.
348Many popular e-mail applications will not always transmit a MIME
349attachment as plain text, making it impossible to comment on your
350code.  A MIME attachment also takes Linus a bit more time to process,
351decreasing the likelihood of your MIME-attached change being accepted.
352
353Exception:  If your mailer is mangling patches then someone may ask
354you to re-send them using MIME.
355
356See Documentation/email-clients.txt for hints about configuring
357your e-mail client so that it sends your patches untouched.
358
3597) E-mail size.
360---------------
361
362Large changes are not appropriate for mailing lists, and some
363maintainers.  If your patch, uncompressed, exceeds 300 kB in size,
364it is preferred that you store your patch on an Internet-accessible
365server, and provide instead a URL (link) pointing to your patch.  But note
366that if your patch exceeds 300 kB, it almost certainly needs to be broken up
367anyway.
368
3698) Respond to review comments.
370------------------------------
371
372Your patch will almost certainly get comments from reviewers on ways in
373which the patch can be improved.  You must respond to those comments;
374ignoring reviewers is a good way to get ignored in return.  Review comments
375or questions that do not lead to a code change should almost certainly
376bring about a comment or changelog entry so that the next reviewer better
377understands what is going on.
378
379Be sure to tell the reviewers what changes you are making and to thank them
380for their time.  Code review is a tiring and time-consuming process, and
381reviewers sometimes get grumpy.  Even in that case, though, respond
382politely and address the problems they have pointed out.
383
384
3859) Don't get discouraged - or impatient.
386----------------------------------------
387
388After you have submitted your change, be patient and wait.  Reviewers are
389busy people and may not get to your patch right away.
390
391Once upon a time, patches used to disappear into the void without comment,
392but the development process works more smoothly than that now.  You should
393receive comments within a week or so; if that does not happen, make sure
394that you have sent your patches to the right place.  Wait for a minimum of
395one week before resubmitting or pinging reviewers - possibly longer during
396busy times like merge windows.
397
398
39910) Include PATCH in the subject
400--------------------------------
401
402Due to high e-mail traffic to Linus, and to linux-kernel, it is common
403convention to prefix your subject line with [PATCH].  This lets Linus
404and other kernel developers more easily distinguish patches from other
405e-mail discussions.
406
407
408
40911) Sign your work
410------------------
411
412To improve tracking of who did what, especially with patches that can
413percolate to their final resting place in the kernel through several
414layers of maintainers, we've introduced a "sign-off" procedure on
415patches that are being emailed around.
416
417The sign-off is a simple line at the end of the explanation for the
418patch, which certifies that you wrote it or otherwise have the right to
419pass it on as an open-source patch.  The rules are pretty simple: if you
420can certify the below:
421
422        Developer's Certificate of Origin 1.1
423
424        By making a contribution to this project, I certify that:
425
426        (a) The contribution was created in whole or in part by me and I
427            have the right to submit it under the open source license
428            indicated in the file; or
429
430        (b) The contribution is based upon previous work that, to the best
431            of my knowledge, is covered under an appropriate open source
432            license and I have the right under that license to submit that
433            work with modifications, whether created in whole or in part
434            by me, under the same open source license (unless I am
435            permitted to submit under a different license), as indicated
436            in the file; or
437
438        (c) The contribution was provided directly to me by some other
439            person who certified (a), (b) or (c) and I have not modified
440            it.
441
442        (d) I understand and agree that this project and the contribution
443            are public and that a record of the contribution (including all
444            personal information I submit with it, including my sign-off) is
445            maintained indefinitely and may be redistributed consistent with
446            this project or the open source license(s) involved.
447
448then you just add a line saying
449
450	Signed-off-by: Random J Developer <random@developer.example.org>
451
452using your real name (sorry, no pseudonyms or anonymous contributions.)
453
454Some people also put extra tags at the end.  They'll just be ignored for
455now, but you can do this to mark internal company procedures or just
456point out some special detail about the sign-off.
457
458If you are a subsystem or branch maintainer, sometimes you need to slightly
459modify patches you receive in order to merge them, because the code is not
460exactly the same in your tree and the submitters'. If you stick strictly to
461rule (c), you should ask the submitter to rediff, but this is a totally
462counter-productive waste of time and energy. Rule (b) allows you to adjust
463the code, but then it is very impolite to change one submitter's code and
464make him endorse your bugs. To solve this problem, it is recommended that
465you add a line between the last Signed-off-by header and yours, indicating
466the nature of your changes. While there is nothing mandatory about this, it
467seems like prepending the description with your mail and/or name, all
468enclosed in square brackets, is noticeable enough to make it obvious that
469you are responsible for last-minute changes. Example :
470
471	Signed-off-by: Random J Developer <random@developer.example.org>
472	[lucky@maintainer.example.org: struct foo moved from foo.c to foo.h]
473	Signed-off-by: Lucky K Maintainer <lucky@maintainer.example.org>
474
475This practice is particularly helpful if you maintain a stable branch and
476want at the same time to credit the author, track changes, merge the fix,
477and protect the submitter from complaints. Note that under no circumstances
478can you change the author's identity (the From header), as it is the one
479which appears in the changelog.
480
481Special note to back-porters: It seems to be a common and useful practice
482to insert an indication of the origin of a patch at the top of the commit
483message (just after the subject line) to facilitate tracking. For instance,
484here's what we see in a 3.x-stable release:
485
486Date:   Tue Oct 7 07:26:38 2014 -0400
487
488    libata: Un-break ATA blacklist
489
490    commit 1c40279960bcd7d52dbdf1d466b20d24b99176c8 upstream.
491
492And here's what might appear in an older kernel once a patch is backported:
493
494    Date:   Tue May 13 22:12:27 2008 +0200
495
496        wireless, airo: waitbusy() won't delay
497
498        [backport of 2.6 commit b7acbdfbd1f277c1eb23f344f899cfa4cd0bf36a]
499
500Whatever the format, this information provides a valuable help to people
501tracking your trees, and to people trying to troubleshoot bugs in your
502tree.
503
504
50512) When to use Acked-by: and Cc:
506---------------------------------
507
508The Signed-off-by: tag indicates that the signer was involved in the
509development of the patch, or that he/she was in the patch's delivery path.
510
511If a person was not directly involved in the preparation or handling of a
512patch but wishes to signify and record their approval of it then they can
513ask to have an Acked-by: line added to the patch's changelog.
514
515Acked-by: is often used by the maintainer of the affected code when that
516maintainer neither contributed to nor forwarded the patch.
517
518Acked-by: is not as formal as Signed-off-by:.  It is a record that the acker
519has at least reviewed the patch and has indicated acceptance.  Hence patch
520mergers will sometimes manually convert an acker's "yep, looks good to me"
521into an Acked-by: (but note that it is usually better to ask for an
522explicit ack).
523
524Acked-by: does not necessarily indicate acknowledgement of the entire patch.
525For example, if a patch affects multiple subsystems and has an Acked-by: from
526one subsystem maintainer then this usually indicates acknowledgement of just
527the part which affects that maintainer's code.  Judgement should be used here.
528When in doubt people should refer to the original discussion in the mailing
529list archives.
530
531If a person has had the opportunity to comment on a patch, but has not
532provided such comments, you may optionally add a "Cc:" tag to the patch.
533This is the only tag which might be added without an explicit action by the
534person it names - but it should indicate that this person was copied on the
535patch.  This tag documents that potentially interested parties
536have been included in the discussion.
537
538
53913) Using Reported-by:, Tested-by:, Reviewed-by:, Suggested-by: and Fixes:
540--------------------------------------------------------------------------
541
542The Reported-by tag gives credit to people who find bugs and report them and it
543hopefully inspires them to help us again in the future.  Please note that if
544the bug was reported in private, then ask for permission first before using the
545Reported-by tag.
546
547A Tested-by: tag indicates that the patch has been successfully tested (in
548some environment) by the person named.  This tag informs maintainers that
549some testing has been performed, provides a means to locate testers for
550future patches, and ensures credit for the testers.
551
552Reviewed-by:, instead, indicates that the patch has been reviewed and found
553acceptable according to the Reviewer's Statement:
554
555	Reviewer's statement of oversight
556
557	By offering my Reviewed-by: tag, I state that:
558
559 	 (a) I have carried out a technical review of this patch to
560	     evaluate its appropriateness and readiness for inclusion into
561	     the mainline kernel.
562
563	 (b) Any problems, concerns, or questions relating to the patch
564	     have been communicated back to the submitter.  I am satisfied
565	     with the submitter's response to my comments.
566
567	 (c) While there may be things that could be improved with this
568	     submission, I believe that it is, at this time, (1) a
569	     worthwhile modification to the kernel, and (2) free of known
570	     issues which would argue against its inclusion.
571
572	 (d) While I have reviewed the patch and believe it to be sound, I
573	     do not (unless explicitly stated elsewhere) make any
574	     warranties or guarantees that it will achieve its stated
575	     purpose or function properly in any given situation.
576
577A Reviewed-by tag is a statement of opinion that the patch is an
578appropriate modification of the kernel without any remaining serious
579technical issues.  Any interested reviewer (who has done the work) can
580offer a Reviewed-by tag for a patch.  This tag serves to give credit to
581reviewers and to inform maintainers of the degree of review which has been
582done on the patch.  Reviewed-by: tags, when supplied by reviewers known to
583understand the subject area and to perform thorough reviews, will normally
584increase the likelihood of your patch getting into the kernel.
585
586A Suggested-by: tag indicates that the patch idea is suggested by the person
587named and ensures credit to the person for the idea. Please note that this
588tag should not be added without the reporter's permission, especially if the
589idea was not posted in a public forum. That said, if we diligently credit our
590idea reporters, they will, hopefully, be inspired to help us again in the
591future.
592
593A Fixes: tag indicates that the patch fixes an issue in a previous commit. It
594is used to make it easy to determine where a bug originated, which can help
595review a bug fix. This tag also assists the stable kernel team in determining
596which stable kernel versions should receive your fix. This is the preferred
597method for indicating a bug fixed by the patch. See #2 above for more details.
598
599
60014) The canonical patch format
601------------------------------
602
603This section describes how the patch itself should be formatted.  Note
604that, if you have your patches stored in a git repository, proper patch
605formatting can be had with "git format-patch".  The tools cannot create
606the necessary text, though, so read the instructions below anyway.
607
608The canonical patch subject line is:
609
610    Subject: [PATCH 001/123] subsystem: summary phrase
611
612The canonical patch message body contains the following:
613
614  - A "from" line specifying the patch author (only needed if the person
615    sending the patch is not the author).
616
617  - An empty line.
618
619  - The body of the explanation, line wrapped at 75 columns, which will
620    be copied to the permanent changelog to describe this patch.
621
622  - The "Signed-off-by:" lines, described above, which will
623    also go in the changelog.
624
625  - A marker line containing simply "---".
626
627  - Any additional comments not suitable for the changelog.
628
629  - The actual patch (diff output).
630
631The Subject line format makes it very easy to sort the emails
632alphabetically by subject line - pretty much any email reader will
633support that - since because the sequence number is zero-padded,
634the numerical and alphabetic sort is the same.
635
636The "subsystem" in the email's Subject should identify which
637area or subsystem of the kernel is being patched.
638
639The "summary phrase" in the email's Subject should concisely
640describe the patch which that email contains.  The "summary
641phrase" should not be a filename.  Do not use the same "summary
642phrase" for every patch in a whole patch series (where a "patch
643series" is an ordered sequence of multiple, related patches).
644
645Bear in mind that the "summary phrase" of your email becomes a
646globally-unique identifier for that patch.  It propagates all the way
647into the git changelog.  The "summary phrase" may later be used in
648developer discussions which refer to the patch.  People will want to
649google for the "summary phrase" to read discussion regarding that
650patch.  It will also be the only thing that people may quickly see
651when, two or three months later, they are going through perhaps
652thousands of patches using tools such as "gitk" or "git log
653--oneline".
654
655For these reasons, the "summary" must be no more than 70-75
656characters, and it must describe both what the patch changes, as well
657as why the patch might be necessary.  It is challenging to be both
658succinct and descriptive, but that is what a well-written summary
659should do.
660
661The "summary phrase" may be prefixed by tags enclosed in square
662brackets: "Subject: [PATCH <tag>...] <summary phrase>".  The tags are
663not considered part of the summary phrase, but describe how the patch
664should be treated.  Common tags might include a version descriptor if
665the multiple versions of the patch have been sent out in response to
666comments (i.e., "v1, v2, v3"), or "RFC" to indicate a request for
667comments.  If there are four patches in a patch series the individual
668patches may be numbered like this: 1/4, 2/4, 3/4, 4/4.  This assures
669that developers understand the order in which the patches should be
670applied and that they have reviewed or applied all of the patches in
671the patch series.
672
673A couple of example Subjects:
674
675    Subject: [PATCH 2/5] ext2: improve scalability of bitmap searching
676    Subject: [PATCH v2 01/27] x86: fix eflags tracking
677
678The "from" line must be the very first line in the message body,
679and has the form:
680
681        From: Original Author <author@example.com>
682
683The "from" line specifies who will be credited as the author of the
684patch in the permanent changelog.  If the "from" line is missing,
685then the "From:" line from the email header will be used to determine
686the patch author in the changelog.
687
688The explanation body will be committed to the permanent source
689changelog, so should make sense to a competent reader who has long
690since forgotten the immediate details of the discussion that might
691have led to this patch.  Including symptoms of the failure which the
692patch addresses (kernel log messages, oops messages, etc.) is
693especially useful for people who might be searching the commit logs
694looking for the applicable patch.  If a patch fixes a compile failure,
695it may not be necessary to include _all_ of the compile failures; just
696enough that it is likely that someone searching for the patch can find
697it.  As in the "summary phrase", it is important to be both succinct as
698well as descriptive.
699
700The "---" marker line serves the essential purpose of marking for patch
701handling tools where the changelog message ends.
702
703One good use for the additional comments after the "---" marker is for
704a diffstat, to show what files have changed, and the number of
705inserted and deleted lines per file.  A diffstat is especially useful
706on bigger patches.  Other comments relevant only to the moment or the
707maintainer, not suitable for the permanent changelog, should also go
708here.  A good example of such comments might be "patch changelogs"
709which describe what has changed between the v1 and v2 version of the
710patch.
711
712If you are going to include a diffstat after the "---" marker, please
713use diffstat options "-p 1 -w 70" so that filenames are listed from
714the top of the kernel source tree and don't use too much horizontal
715space (easily fit in 80 columns, maybe with some indentation).  (git
716generates appropriate diffstats by default.)
717
718See more details on the proper patch format in the following
719references.
720
72115) Explicit In-Reply-To headers
722--------------------------------
723
724It can be helpful to manually add In-Reply-To: headers to a patch
725(e.g., when using "git send-email") to associate the patch with
726previous relevant discussion, e.g. to link a bug fix to the email with
727the bug report.  However, for a multi-patch series, it is generally
728best to avoid using In-Reply-To: to link to older versions of the
729series.  This way multiple versions of the patch don't become an
730unmanageable forest of references in email clients.  If a link is
731helpful, you can use the https://lkml.kernel.org/ redirector (e.g., in
732the cover email text) to link to an earlier version of the patch series.
733
734
73516) Sending "git pull" requests
736-------------------------------
737
738If you have a series of patches, it may be most convenient to have the
739maintainer pull them directly into the subsystem repository with a
740"git pull" operation.  Note, however, that pulling patches from a developer
741requires a higher degree of trust than taking patches from a mailing list.
742As a result, many subsystem maintainers are reluctant to take pull
743requests, especially from new, unknown developers.  If in doubt you can use
744the pull request as the cover letter for a normal posting of the patch
745series, giving the maintainer the option of using either.
746
747A pull request should have [GIT] or [PULL] in the subject line.  The
748request itself should include the repository name and the branch of
749interest on a single line; it should look something like:
750
751  Please pull from
752
753      git://jdelvare.pck.nerim.net/jdelvare-2.6 i2c-for-linus
754
755  to get these changes:
756
757A pull request should also include an overall message saying what will be
758included in the request, a "git shortlog" listing of the patches
759themselves, and a diffstat showing the overall effect of the patch series.
760The easiest way to get all this information together is, of course, to let
761git do it for you with the "git request-pull" command.
762
763Some maintainers (including Linus) want to see pull requests from signed
764commits; that increases their confidence that the request actually came
765from you.  Linus, in particular, will not pull from public hosting sites
766like GitHub in the absence of a signed tag.
767
768The first step toward creating such tags is to make a GNUPG key and get it
769signed by one or more core kernel developers.  This step can be hard for
770new developers, but there is no way around it.  Attending conferences can
771be a good way to find developers who can sign your key.
772
773Once you have prepared a patch series in git that you wish to have somebody
774pull, create a signed tag with "git tag -s".  This will create a new tag
775identifying the last commit in the series and containing a signature
776created with your private key.  You will also have the opportunity to add a
777changelog-style message to the tag; this is an ideal place to describe the
778effects of the pull request as a whole.
779
780If the tree the maintainer will be pulling from is not the repository you
781are working from, don't forget to push the signed tag explicitly to the
782public tree.
783
784When generating your pull request, use the signed tag as the target.  A
785command like this will do the trick:
786
787  git request-pull master git://my.public.tree/linux.git my-signed-tag
788
789
790----------------------
791SECTION 2 - REFERENCES
792----------------------
793
794Andrew Morton, "The perfect patch" (tpp).
795  <http://www.ozlabs.org/~akpm/stuff/tpp.txt>
796
797Jeff Garzik, "Linux kernel patch submission format".
798  <http://linux.yyz.us/patch-format.html>
799
800Greg Kroah-Hartman, "How to piss off a kernel subsystem maintainer".
801  <http://www.kroah.com/log/linux/maintainer.html>
802  <http://www.kroah.com/log/linux/maintainer-02.html>
803  <http://www.kroah.com/log/linux/maintainer-03.html>
804  <http://www.kroah.com/log/linux/maintainer-04.html>
805  <http://www.kroah.com/log/linux/maintainer-05.html>
806  <http://www.kroah.com/log/linux/maintainer-06.html>
807
808NO!!!! No more huge patch bombs to linux-kernel@vger.kernel.org people!
809  <https://lkml.org/lkml/2005/7/11/336>
810
811Kernel Documentation/CodingStyle:
812  <Documentation/CodingStyle>
813
814Linus Torvalds's mail on the canonical patch format:
815  <http://lkml.org/lkml/2005/4/7/183>
816
817Andi Kleen, "On submitting kernel patches"
818  Some strategies to get difficult or controversial changes in.
819  http://halobates.de/on-submitting-patches.pdf
820
821--
822